Page History
Table of Contents
Ensemble of Data Assimilations (EDA)
An Ensemble of Data Assimilations(EDA) is an ensemble of independent 4D-Var data assimilations which aims to:
...
Differences between pairs of analyses (and forecast) fields have the statistical characteristics of analysis (and forecast) error.
Fig5.1.1-1: An idealized schematic showing how the 12 hour assimilation window used by 4D-Var (left part of the diagram) modifies the initial trajectories of the members of the ensemble of data assimilations EDA (in blue) to reflect the information from the assimilated observations (black dots with error bars). The analysis trajectories (in green) have taken into account the new observations and thus are confined within a narrower ensemble. Assimilating the new observations reduces the spread. Also a bias has been corrected by reducing the magnitude of some of the largest values in the original ensemble.
...
- the variability of the perturbations does not grow sufficiently through the forecast (perturbations are currently under-dispersive).
Scale-selective EDA re-centring
Generally small-scale structures come directly from the EDA but there were some problems with larger scale structures. For theses features, scale-selective EDA re-centring introduced in Cy50r1 improves the realism of initial conditions, particularly for tropical cyclones. Scale-selective EDA re-centring is only applied to large-scale upper-air fields, centring them on the control forecast. This helps avoid unrealistic 'double-centred' tropical cyclones from appearing at the start of the ENS forecasts.
Fig5.1.1-2: Mean sea level pressure ENS members 36 and 45 at forecast step T+0 h for Tropical Cyclone Freddy 00UTC 17 Feb 2023. Top panels show results with EDA re-centring (Cycle 49r1) producing unrealistic double centre; bottom panels use scale-selective EDA re-centring (Cycle 50r1) and unrealistic 'double centre' feature doesn't appear.
Additional Sources of Information
(Note: In older material there may be references to issues that have subsequently been addressed)
...
Overview
Community Forums
Content Tools